POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES

advertisement
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Abstract. We study some systems of polynomials whose support lies in the convex hull
of a circuit, giving a sharp upper bound for their numbers of real solutions. This upper
bound is non-trivial in that it is smaller than either the Kouchnirenko or the Khovanskii
bounds for these systems. When the support is exactly a circuit whose affine span is
Zn , this bound is 2n + 1, while the Khovanskii bound is exponential in n2 . The bound
2n + 1 can be attained only for non-degenerate circuits. Our methods involve a mixture
of combinatorics, geometry, and arithmetic.
Introduction
The notion of degree for a multivariate Laurent polynomial f is captured by its support,
which is the set of exponent vectors of monomials in f . For example, Kouchnirenko [6]
generalized the classical Bézout Theorem, showing that the number of solutions in the
complex torus (C∗ )n to a generic system of n polynomials in n variables with common
support A is the volume v(A) of the convex hull of A, normalized so that the unit cube
[0, 1]n has volume n!. This Kouchnirenko number is also a (trivial) bound on the number
of real solutions to such a system of real polynomials. This bound is reached, for example,
when the common support A admits a regular unimodular triangulation [8].
Khovanskii [5] gave a bound on the number of real solutions to a polynomial system
that depends only upon the cardinality |A| of its support A:
|A|−1
Number of real solutions ≤ 2n 2( 2 ) · (n + 1)|A|−1 .
This enormous fewnomial bound is non-trivial (smaller than the Kouchnirenko bound)
only when the cardinality of A is very small when compared to the volume of its convex
hull. It is widely believed that significantly smaller bounds should hold. For example, Li,
Rojas, and Wang [7] showed that two trinomials in 2 variables have at most 20 common
solutions, which is much less than the Khovanskii bound of 20736. While significantly
lower bounds are expected, we know of no reasonable conjectures about the nature of
hypothetical lower bounds.
There are few other examples of non-trivial polynomial systems for which it is known
that not all solutions can be real via a bound smaller than the Khovanskii bound. We
describe a class of supports and prove an upper bound (which is often sharp) for the
Part of work done at MSRI was supported by NSF grant DMS-9810361.
Work of Sottile is supported by the Clay Mathematical Institute.
Sottile and Bihan were supported in part by NSF CAREER grant DMS-0134860.
Bertrand is supported by the European research network IHP-RAAG contract HPRN-CT-2001-00271.
1
2
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
number of real solutions to a polynomial system with those supports that is non-trivial
in that it is smaller than either the Kouchnirenko or the Khovanskii bound.
A finite subset A of Zn that affinely spans Zn is primitive. A (possibly degenerate)
circuit is a collection C := {0, w0 , w1 , . . . , wn } ⊂ Zn of n+2 integer vectors which spans
Rn . Here is the simplest version of our main results, which are proven in Sections 4 and 5.
Theorem. A polynomial system with support a primitive circuit has at most 2n + 1
real solutions. There exist systems with support a primitive circuit having 2n + 1 nondegenerate real solutions.
This sharp bound for circuits suggests that one may optimistically expect similar dramatic improvements in the doubly exponential Khovanski bound for other sets A of supports.
Adding the vectors 2w0 , 3w0 , . . . , kw0 to a circuit C produces a near circuit if Rw0 ∩ C =
{0, w0 }. Let ν be the cardinality of D ∩ {w1 , . . . , wn }, where D ⊂ C is a minimal affinely
dependent subset. Let ℓ be the largest integer so that w0 /ℓ is integral. We show that
a polynomial system with support a primitive near circuit has at most k(2ν−1) + 2 real
solutions if ℓ is odd, or at most 2kν + 1 real solutions if ℓ is even, and these bounds
are tight among all such near circuits. These bounds coincide when k = 1, that is, for
primitive circuits. This is proven in Theorem 5.6.
For a given near circuit, we use its geometry and arithmetic to give tighter upper bounds
and construct systems with many real zeroes.
An important step is to determine an eliminant for the system, which has the form
p
ν
Y
Y
N
ℓ λi
(∗)
x
(gi (x )) −
(gi (xℓ ))λi ,
i=1
i=p+1
where the polynomials gi all have degree k, and N and λi are the coefficients of the
minimal linear dependence relation among D ∪ {w0 /ℓ}. This reduces the problem to
studying the possible numbers of real zeroes of such a univariate polynomial. We adapt a
method of Khovanskii to establish an upper bound for the number of real zeroes of such a
polynomial. Our upper bound uses a variant of the Viro construction, which also allows
us to construct polynomials (∗) with many real zeroes.
In Section 1, we establish some basics on sparse polynomial systems, and then devote
Section 2 to an example of a family of near circuits for which it is easy to establish sharp
upper bounds, as simple linear algebra suffices for the elimination, Descartes’s rule of signs
gives the bounds, and the Viro construction establishes their sharpness. In Section 3, we
compute an eliminant of the form (∗) for a system supported on a near circuit. Section 4
is devoted to proving an upper bound for the number of real solutions to a polynomial
of the form (∗), while in Section 5 we construct such polynomials with many real zeroes,
and describe some cases in which our bounds are sharp, including when A is a circuit.
The authors wish to thank Stepan Orevkov and Andrei Gabrielov for useful discussions.
1. Basics on sparse polynomial systems
We emphasize that we look for solutions to polynomial systems which have only non-zero
coordinates and thus lie in (C∗ )n . Write xw for the monomial with exponent vector w ∈ Zn .
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
3
Let A ⊂ Zn be a finite set which does not lie in an affine hyperplane. A polynomial f has
support A if the exponent vectors of its monomials lie in A. A polynomial system with
support A is a system
(1.1)
f1 (x1 , . . . , xn ) = f2 (x1 , . . . , xn ) = · · · = fn (x1 , . . . , xn ) = 0 ,
where each polynomial fi has support A. Multiplying a polynomial f by a monomial xw
does not change its set of zeroes in (C∗ )n , but does translate its support by the vector
w. Thus it is no loss to assume that 0 ∈ A. A system (1.1) is generic if its number of
solutions in (C∗ )n equals v(A), the volume of the convex hull of A normalized so that
the unit cube [0, 1]n has volume n!, which is the Kouchnirenko bound [6]. This condition
forces each solution to be simple. We will always assume that our systems are generic in
this sense.
Let ZA ⊂ Zn be the full rank sublattice generated by the vectors in A. (This is the
affine span of A as 0 ∈ A.) If ZA = Zn , then A is primitive. The index of A is the index
of ZA in Zn . The fundamental theorem of abelian groups implies that
Zn
Z
Z
Z
≃
⊕
⊕ ··· ⊕
,
ZA
a1 Z a2 Z
an Z
where ai divides ai+1 , for i = 1, . . . , n−1. These numbers a1 , . . . , an are the invariant
factors of A. When A is a simplex—so there are n non-zero vectors in A—then the
numbers ai are the invariant factors of the matrix whose columns are these vectors. The
index of A is the product of its invariant factors.
1.1. Polynomial systems with support a simplex. Let e(A) be the number of even
invariant factors of A. The following result can be found in Section 3 of [8].
Proposition 1.1. Suppose that A is the set of vertices of a simplex. Then the number of
real solutions to a generic system with support A is
(i) 0 or 2e(A) if v(A) is even.
(ii) 1 if v(A) is odd.
Proof. Suppose that 0 ∈ A. Given a polynomial system (1.1) with support A whose coefficients are generic, we may perform Gaussian elimination on the matrix of its coefficients
and convert it into a system of the form
(1.2)
xwi = βi ,
for i = 1, . . . , n ,
where βi 6= 0 and w1 , . . . , wn are the non-zero elements of A. Solutions to this system
∗ n
have the form ϕ−1
A (β), where β = (β1 , . . . , βn ) ∈ (C ) and ϕA is the homomorphism
(C∗ )n ∋ (x1 , . . . , xn ) 7−→ (xw1 , . . . , xwn ) ∈ (C∗ )n .
Real solutions to (1.2) are ψ −1 (β), where β ∈ (R∗ )n and ψ : (R∗ )n → (R∗ )n is the restriction of ϕA to (R∗ )n . The kernel of ψ consists of those points x ∈ {±1}n that satisfy
xwi = 1 for i = 1, . . . , n .
Let A be a matrix whose columns are the non-zero elements of A. Identifying {±1} with
Z/2Z identifies the kernel of ψ with the kernel of the reduction of A modulo 2, which has
4
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
dimension e(A). The result follows as ψ is surjective if (and only if) e(A) = 0, which is
equivalent to v(A) being odd.
Remark 1.2. The upper bound of Proposition 1.1 for a system with support A is the
volume of A if and only if no invariant factor of A exceeds 2.
1.2. Generic systems. The system (1.2) is generic (has v(A) simple roots in (C∗ )n )
when the numbers βi are non-zero. We give a proof of the following elementary result, as
we will use the proof later.
Proposition 1.3. Suppose that A does not lie in an affine hyperplane. Then there is
a non-empty Zariski open subset in the space of coefficients of monomials appearing in a
system with support A such that the system has v(A) simple solutions in (C∗ )n .
Proof. Suppose that 0 ∈ A. Then the affine span ZA is a full rank sublattice of Zn . Let
a := |A| − 1 and consider the map ϕA defined by
(1.3)
ϕA : (C∗ )n ∋ (x1 , . . . , xn ) 7−→ [xw | w ∈ A] ∈ Pa .
Its image is a subgroup of the dense torus in Pa , and it is a homomorphism of algebraic
groups. Then a polynomial system (1.1) with support A is the pullback of n linear forms
(given by the coefficients of the fi ) along the map ϕA . These n linear forms determine a
linear section of the closure XA of the image of ϕA , a (not necessarily normal) projective
toric variety [9, §4,13]. Bertini’s Theorem [4, p. 179] asserts that a general linear section
of XA consists of deg(XA ) simple points, all lying in the image of ϕA . Each of these pull
back along ϕA to |ker(ϕA )| distinct solutions to the original system.
The kernel of ϕA is the abelian group dual to the factor group Zn /ZA, which has order
equal to the index of A. Furthermore, the toric variety XA has degree equal to the volume
of the convex hull of A, normalized so that a unit parallelepiped of ZA has volume n!.
The product of this volume with the index of A is the usual volume of A. Thus a system
with support A and generic coefficients has v(A) simple solutions.
1.3. Congruences on the number of real solutions. The proof of Proposition 1.3
shows that the solutions to the system (1.1) are the fibres ϕ−1
A (β) for the points β in a
linear section of XA . It follows that Proposition 1.1 gives some restrictions on the possible
numbers of real solutions. The index of A factors as 2e(A) · N .
Corollary 1.4. The number of real solutions to (1.1) is at most v(A)/N and is congruent
to this number modulo max{2, 2e(A) }.
Proof. Let ϕA be the map (1.3). The image of any real solution to (1.1) under ϕA is
a real point, and there are 2e(A) real solutions with the same image under ϕA . Thus
the maximum number of real solutions to (1.1) is 2e(A) · deg XA , which is v(A)/N . The
congruence follows as the restriction of the map ϕA to the real subtorus is surjective on
the real subtorus of XA if and only if e(A) = 0.
Remark 1.5. Any bound or construction for the number of real roots of polynomial
systems associated to primitive vector configurations gives the same bounds and constructions for the number of real roots for configurations with odd index. Indeed, let
A ⊂ Zn be finite and B ⊂ Zn be a basis for ZA so that linear combinations of vectors
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
5
in B identify ZA with Zn . This identification maps A to a primitive vector configuration
A′ , which has the same geometry and arithmetic as A. The map ϕA (1.3) factors
ϕB∪{0}
ϕ
′
(C∗ )n −−−−→ (C∗ )n −−A→ Pa .
Since A and B ∪ {0} have the same (odd) index, ϕB∪{0} is bijective on (R∗ )n . Thus ϕB∪{0}
gives a bijection between real solutions to systems with support the primitive vector
configuration A′ and real solutions to systems with support A.
2. A family of systems with a sharp bound
We describe a family of supports ∆ ⊂ Zn and prove a non-trivial sharp upper bound on
the number of real solutions to polynomial systems with support ∆. That is, the points
in ∆ affinely span Zn , but there are fewer than v(∆) real solutions to polynomial systems
with support ∆. These sets ∆ also have the property that they consist of all the integer
points in their convex hull.
Let l > k > 0 and n ≥ 3 be integers and ǫ = (ǫ1 , . . . , ǫn−1 ) ∈ {0, 1}n−1 be non-zero.
Then ∆ǫk,l ⊂ Rn consists of the points
(0, . . . , 0), (1, 0, . . . , 0), . . . , (0, . . . , 0, 1, 0), (0, . . . , 0, k), (ǫ1 , . . . , ǫn−1 , l) .
together with the points along the last axis
(0, . . . , 0, 1), (0, . . . , 0, 2), . . . , (0, . . . , 0, k−1) .
Since these include
the standard basis and the origin, ∆ǫk,l is primitive.
P
Set |ǫ| := i ǫi . Then the volume of ∆ǫk,l is l + k|ǫ|. Indeed, the configuration ∆ǫk,l can
be triangulated into two simplices ∆ǫk,l \ {(ǫ1 , . . . , ǫn−1 , l)} and ∆ǫk,l \ {0} with volumes k
and l − k + k|ǫ|, respectively. One way to see this is to apply the affine transformation
(x1 , . . . , xn ) 7−→ (x1 , . . . , xn−1 , xn − k + k
n−1
X
xi ) .
i=1
Theorem 2.1. The number, r, of real solutions to a generic system of n real polynomials
with support ∆ǫk,l lies in the interval
0 ≤ r ≤ k + k|ǫ| + 2 ,
and every number in this interval with the same parity as l + k|ǫ| occurs.
This upper bound does not depend on l and, since k < l, it is smaller than or equal to
the number l + k|ǫ| of complex solutions. We use elimination to prove this result.
6
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Example 2.2. Suppose that n = k = 3, l = 5, and ǫ = (1, 1).
xyz 5
Then the system
x + y + xyz 5 + 1 + z + z 2 + z 3 = 0
x + 2y + 3xyz 5 + 5 + 7z + 11z 2 + 13z 3 = 0
2x + 2y + xyz 5 + 4 + 8z + 16z 2 + 32z 3 = 0
is equivalent to
x − (5 + 11z + 23z 2 + 41z 3 ) = 0
y + (8 + 18z + 38z 2 + 72z 3 ) = 0
xyz 5 − (2 + 6z + 14z 2 + 30z 3 ) = 0
z3
z2
z
1
y
x
And thus its number of real roots equals the number of real roots of
z 5 (5 + 11z + 23z 2 + 41z 3 )(8 + 18z + 38z 2 + 72z 3 ) − (2 + 6z + 14z 2 + 30z 3 ) ,
which, as we invite the reader to check, is 3.
Proof. A generic real polynomial system with support ∆ǫk,l has the form
n−1
X
cij xj + cin xǫ xln + fi (xn ) = 0
for i = 1, . . . , n ,
j=1
ǫ
n−1
where each polynomial fi has degree k and xǫ is the monomial xǫ11 · · · xn−1
.
Since all solutions to our system are simple, we may perturb the coefficient matrix
(cij )ni,j=1 if necessary and then use Gaussian elimination to obtain an equivalent system
(2.1)
x1 − g1 (xn ) = · · · = xn−1 − gn−1 (xn ) = xǫ xn l − gn (xn ) = 0 ,
where each polynomial gi has degree k. Using the first n − 1 polynomials to eliminate the
variables x1 , . . . , xn−1 gives the univariate polynomial
(2.2)
xn l · g1 (xn )ǫ1 · · · gn−1 (xn )ǫn−1 − gn (xn ) ,
which has degree l + k|ǫ| = v(∆ǫk,l ). Any zero of this polynomial leads to a solution of the
original system (2.1) by back substitution. This implies that the number of real roots of
the polynomial (2.2) is equal to the number of real solutions to our original system (2.1).
The eliminant (2.2) has no terms of degree m for k < m < l, and so it has at most
k + k|ǫ| + 2 non-zero real roots, by Descartes’s rule of signs ([3, 1], see also Remark 2.3).
This proves the upper bound. We complete the proof by constructing a polynomial of the
form (2.2) having r roots, for every number r in the interval between 0 and k + k|ǫ| + 2
with the same parity as l + k|ǫ|.
Choose real polynomials f1 , . . . , fn of degree k having simple roots and non-zero constant terms. We further assume that the roots of f1 , . . . , fn−1 are distinct. Put f ǫ =
ǫn−1
f1ǫ1 · · · fn−1
and let α be the ratio of the leading term of fn to the constant term of f ǫ .
Choose any piecewise-linear convex function ν : [0, k|ǫ| + l] → R which is identically 0 on
[l, k|ǫ| + l] and whose maximal domains of linearity are [0, k], [k, l], and [l, k|ǫ| + l].
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
7
Let f (z) := fn ± αz l f ǫ (the sign ± will be determined later). The Viro polynomial ft
associated to f and ν is obtained by multiplying the monomial ap z p in f (z) by tν(p) ([10,
ǫn−1
11, 2]). By the definition of ν, ft = fn,t +αz l f1ǫ1 · · · fn−1
, where fn,t is the Viro polynomial
obtained from fn and the restriction of ν to [0, k]. By Viro’s Theorem ([10, 11, 2], see
also Proposition 4.1), there exists a sufficiently small t0 > 0, such that if t0 > t > 0, then
the polynomial ft will have r = r1 + r2 + r3 simple real roots, where r1 is the number of
real roots of f ǫ , r2 the number of real roots of fn , and r3 is the number of (non-zero) real
roots of the binomial obtained as the truncation of f to the interval [k, l]. By our choice
of α, this binomial is a constant multiple of z k ± z l . Thus r3 is 1 if l − k is odd, and
either 0 or 2 (depending on the sign ±) if l − k is even. If k is even, then every possible
value of r between 0 and k + k|ǫ| + 2 with the same parity as l + k|ǫ| can be obtained in
this way. If k is odd then this construction gives all admissible values of r in the interval
[|ǫ| + 2, k + k|ǫ| + 2] but no values of r less than |ǫ| + 2.
Suppose now that k is odd. Take one of the Viro polynomials ft with r′ ≥ |ǫ| + 2 real
roots. For a generic t, the crital values of ft are all different. Choose one such t and
suppose without loss of generality that the leading coefficient of ft is positive. For each λ,
consider the polynomial hλ = −λ − ft . If λ is larger than every critical value of ft , then
hλ has either 0 or 1 real roots, depending upon the parity of k|ǫ| + l. Since the number
of real roots of hλ changes by 2 when λ passes through critical values of ft , and h0 has at
least |ǫ| + 2 real roots, every possible number of real roots between 0 and |ǫ| + 2 having
the same parity as l + k|ǫ| occurs for some hλ .
Pd
pi
Remark 2.3. [Descartes’s bound] Let f (x) =
be a univariate polynomial
i=1 ai x
with exponents p1 < · · · < pd and a1 , . . . , ad are non-zero real numbers. Descartes’s rule
of signs asserts that the number of positive roots of f is no more than the number of
i ∈ {1, . . . , d − 1} with ai · ai+1 < 0. Applying this to f (x) and f (−x) shows that the
Pd−1
(pi+1 − pi ), where a = 1 or 2
number of non-zero real roots of f is no more than i=1
according as a is odd or even, respectively.
3. Elimination for near circuits
We first consider (possibly degenerate) circuits, which are collections of n+2 integer
vectors that affinely span Rn .
3.1. Arithmetic of circuits. Suppose that C := {w−1 , w0 , w1 , . . . , wn } ⊂ Zn affinely
spans Rn . For each i = −1, 0, . . . , n, let Ai be the (possibly degenerate) simplex with
vertices C \ {wi }. For any j ∈ {−1, 0, . . . , n}, Cramer’s rule implies that
j−1
X
i
(−1) det(Wi )(wi − wj ) =
i=−1
n
X
(−1)i det(Wi )(wi − wj ) ,
i=j+1
where Wi is the matrix whose columns are the vectors w−1 − wj , . . . , wn − wj , with wi − wj
and wj − wj omitted. Thus if i 6= j, | det Wi | is the volume v(Ai ) of Ai .
8
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Lemma 3.1. Suppose that {0, v0 , . . . , vn } ⊂ Zn is primitive with primitive relation
n
X
αi vi = 0 .
i=0
If Aq := {0, v0 , . . . , vbq , . . . , vn }, then Zn /ZAq ≃ Z/αq Z.
Proof. We can assume that q = 0. Since {0, v0 , . . . , vn } is primitive, Zn = Z{v0 , v1 , . . . , vn }.
Then the image of v0 generates the factor group Zn /Z{v1 , . . . , vn }, and so this factor group
is cyclic. Since the relation is primitive, |α0 | is the least positive multiple of v0 lying in
the lattice Z{v1 , . . . , vn }, which implies that Zn /Z{v1 , . . . , vn } ≃ Z/α0 Z.
When C = {w−1 , w0 , w1 , . . . , wn } is primitive, this proof shows that the invariant factors
of Aq are 1, . . . , 1, |αq |, and v(Aq ) = |αq |. In general, the index a of C is the greatest
common divisor of the volumes v(Ai ), and the primitive affine relation on C has the form
n
X
αi wi = 0
n
X
with
i=−1
αi = 0 ,
i=−1
where a|αi | = v(Ai ). Note that the configuration C admits two triangulations. One is
given by those Ai with αi > 0 and the other by those Ai with αi < 0.
From now on, we make the following assumptions. First, assume w−1 = 0 and choose
signs so that α0 ≥ 0. Furthermore, assume the vectors w1 , . . . , wn are ordered so that
α1 , . . . , αp > 0, αp+1 , . . . , αν < 0, and αν+1 , . . . , αn = 0, for some integers 0 ≤ p ≤ ν ≤ n.
If a is the index of C and we write λi = |αi | = v(Ai )/a, then the primitive relation on C is
p
X
λ i wi =
i=0
and we have
(3.1)
v(C) = a · max
ν
X
λ i wi ,
i=p+1
( p
X
i=0
λi ,
ν
X
i=p+1
λi
)
,
as at least one of {A0 , . . . , Ap } or {Ap+1 , . . . , Aν } is a triangulation of C. If A−1 is nondegenerate, then adding it to the non-triangulation produces a triangulation. Thus v(A−1 )
is the difference of the two numbers in (3.1).
3.2. Systems with support a near circuit. Let {0, w0 , w1 , . . . , wn } ⊂ Zn span Rn and
suppose that w0 = ℓen , where en is the nth standard basis vector and ℓ is a positive
integer. Let k > 0 and consider a generic polynomial system with support
(3.2)
C := {0, w0 , 2w0 , . . . , kw0 , w1 , . . . , wn } .
We will call such a set of vectors a near circuit if no wi for i > 0 lies in Rw0 . We will assume
that C is primitive, which implies that {0, en , w1 , . . . , wn } is primitive. By Remark 1.5,
this will enable us to deduce results for near circuits having odd index.
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
9
Write each vector wi = vi + li · en , where 0 6= vi ∈ Zn−1 . Then {0, v1 , . . . , vn } is
primitive. Let λ1 , . . . , λν be the positive integral coefficients in the primitive relation on
{0, v1 , . . . , vn },
p
ν
X
X
λi vi =
λi v i .
i=1
i=p+1
Here, we could have p = 0 or p = ν, so that one of the two sums collapses to 0.
Assume that the vectors are ordered so that
p
ν
X
X
λ i wi −
λ i wi = 0
(3.3)
N en +
i=1
i=p+1
is the primitive relation on {0, en , w1 , . . . , wn }, where
N :=
ν
X
λi li −
p
X
λi li ≥ 0 .
i=1
i=p+1
Consider a generic real polynomial system with support the near circuit C. Perturbing
the matrix of coefficients of the monomials xwi for i = 1, . . . , n and applying Gaussian
elimination gives a system with the same number of real solutions, but of the form
(3.4)
xwi = gi (xℓn )
for i = 1, . . . , n ,
where each gi is a generic polynomial of degree k.
Define a polynomial f ∈ R[xn ] by
(3.5)
f (xn ) :=
xN
n
p
ν
Y
Y
ℓ λi
(gi (xn )) −
(gi (xℓn ))λi .
i=1
i=p+1
Here, empty products are equal to 1. By (3.1), the degree of f is equal to the volume of
the near circuit C, as N is the volume of A0 and kℓλi is the volume of the simplex Ai ,
which is the convex hull of {kw0 = kℓen , 0, w1 , . . . , wbi , . . . , wn }. Lastly, the absolute value
of the difference
p
ν
X
X
kℓλi −
kℓλi
(3.6)
δ := N +
i=1
i=p+1
in the degrees of the two terms of f is the volume of the convex hull of {kℓen , w1 , . . . , wn },
which may be zero.
Theorem 3.2. Assume that g1 , . . . , gn are generic polynomials of degree k. Then, the association of a solution x of (3.4) to its nth coordinate xn gives a one-to-one correspondence
between solutions of (3.4) and roots of the univariate polynomial f (3.5) which restricts to
a bijection between real solutions to (3.4) and real roots of f .
In particular, bounds on the number of real roots of polynomials of the form (3.5) give
bounds on the number of real solutions to a generic polynomial system with support C.
The polynomial f is the eliminant of the system (3.4), but our proof is not as direct as
the corresponding proof in Section 2.
10
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Lemma 3.3. For each q = 1, . . . , ν, the system
½
xwi = gi (xℓn ),
for i = 1, . . . , n, i 6= q
Iq :
f (xn ) = 0
is equivalent to the system
½
xwi = gi (xℓn ),
for i = 1, . . . , n, i 6= q
¡
¢λq
¡ w ¢λq
Jq :
ℓ
= gq (xn )
x q
Proof. This follows from (3.3) and the form (3.5) of f .
Remark 3.4. Observe that the last polynomial in the system Jq factors
Y¡
¢
(xwq )λq − (gq (xℓn ))λq =
ζxwq − gq (xℓn ) ,
ζ∈Zλq
where Zλq is the set of roots of z λq − 1. For a solution x to Jq , the number ζ is x−wq gq (xℓn ).
This factorization reveals that the system Jq is a disjunction of λq systems with support
C. The subsystem with ζ = 1 is our original system (3.4).
Proof of Theorem 3.2. Since the integers λ1 , . . . , λν are coprime, at least one λi is odd.
We restrict ourselves to the case where λ1 is odd since the proof with any other λi odd is
similar. Let q = 1 in Lemma 3.3 and let xn be a root of f . We show that xn extends to a
unique solution of (3.4), and that the solution is real if and only if xn is real.
We first prolong xn to solutions to the system I1 by solving the system for y ∈ (C∗ )n−1
(3.7)
ℓ
i
y vi = x−l
n gi (xn )
for i = 2, . . . , n .
ℓ
i
The numbers βi := x−l
n gi (xn ) are well-defined and non-zero since the gi are generic polynomials of degree k. Hence (3.7) is a system associated to the simplex B = {0, v2 , . . . , vn } ⊂
∗ n−1
Zn−1 as studied in Proposition 1.1. Its set of solutions is ϕ−1
→
B (β), where ϕB : (C )
(C∗ )n−1 is the homomorphism
ϕB : t 7−→ (tv2 , . . . , tvn ) ∈ (C∗ )n−1 .
−1
Let y ∈ ϕ−1
B (β). Then the fibre ϕB (β) is the set {ty | t ∈ ker(ϕB )}. We determine
which of these, if any, is a solution to (3.4) by computing the number ζ of Remark 3.4.
−v1 −v1 −l1
For ty ∈ ϕ−1
y xn g1 (xℓn ). Consider the map
B (β), this number is t
ψ : ker(ϕB ) ∋ t 7−→ tv1 ∈ Zλ1 .
Since D = {0, v1 , . . . , vn } ⊂ Zn−1 is primitive, ψ is well-defined by Lemma 3.1. Similarly,
ϕD is injective (see Proposition 1.3), and thus so is ψ. Note that ker(ϕB ) is isomorphic
to Zλ1 , thus ψ is an isomorphism. Thus exactly one of these prolongations of xn given by
ϕ−1
B (β) is a solution to (3.4).
Suppose now that xn is real. Then β is real. Since λ1 is odd, there is a unique real
ℓ
1
solution y to (3.7), by Proposition 1.1. But then y −v1 x−l
n g1 (xn ) is real. As λ1 is odd,
there is eactly one real λ1 -th root of unity, namely 1, which proves that the real solution
(y, xn ) is a real solution to (3.4). This completes the proof of the theorem.
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
11
Remark 3.5. The primitivity of C implies that N and ℓ are coprime if N 6= 0, or ℓ = 1 if
N = 0. Indeed, the affine span of C is equal to that of {ℓen , w1 , . . . , wn } and the Cramer
relation on this set is obtained by multiplying both sides of (3.3) by ℓ. Hence, the index
of C is the greatest common divisor of N, ℓλ1 , . . . , ℓλν , and the result follows as the λi are
coprime. In particular, either ℓ is odd, or ℓ is even and N and δ are odd.
Example 3.6. We show that any positive integers p, ℓ, N , with N, ℓ coprime if N 6= 0
or ℓ = 1 if N = 0, and any positive coprime integers λ1 , . . . , λν with ν ≤ n correspond to
a primitive near circuit, when one λi = 1. Thus any polynomial of the form (3.5) is the
eliminant of a system with support a primitive near circuit, when one exponent λi = 1.
Assume without loss of generality that λν = 1. Let e1 , . . . , en be the standard basis in
Rn . Let vi := ei for i = 1, . . . , ν − 1. Set
vν :=
p
X
λi v i −
i=1
ν−1
X
λi v i .
i=p+1
P
Since λ1 , . . . , λν are coprime, there exist integers l1 , . . . , lν such that N = νi=p+1 λi li −
Pp
i=1 λi li . If we set wi := vi + li en for i = 1, . . . , ν and wi := ei−1 for i = ν + 1, . . . , n, then
we obtain the relation (3.3) among en , w1 , . . . , wn . It is then easy to see that the near
circuit {0, ℓen , . . . , kℓen , w1 , . . . , wn } is primitive (for any integer k) if we assume that N
and ℓ are coprime, or ℓ = 1 if N = 0.
4. Upper bounds for near circuits
We first give a version of Viro’s construction for univariate polynomials that takes
multiplicities into account. We then use this to establish upper bounds for the number of
real roots of polynomials of the form (3.5) by studying the total variation in the number
of real roots of a pertubation of the eliminant.
4.1. Viro univariate polynomials. Consider a univariate Viro polynomial
ft (y) =
d
X
φp (t) y p ,
p=p0
P
where t is a positive real number, and each coefficient φp (t) is a finite sum q∈Ip cp,q tq
with cp,q ∈ R and q a rational number. Write f for the function of y and t defined by ft .
Let P be the convex hull of the points (p, q) for p0 ≤ p ≤ d and q ∈ Ip . Assume that
P has dimension 2. Its lower hull L is the union of the edges L1 , . . . , Ll of P whose inner
normals have positive second coordinate. Let Ii be the image of Li under the projection
to the first axis. Then the intervals I1 , . . . , Il subdivide the Newton segment [p0 , d] of ft .
Let f (i) be the facial subpolynomial of f for the face Li . That is, f (i) is the sum of
terms cp,q y p such that (p, q) ∈ Li . Suppose that Li is the graph of y 7→ ai y + bi over Ii ,
Expanding ft (yt−ai )/tbi in powers of t gives
(4.1)
ft (yt−ai )
= f (i) (y) + tAi d(i) (y) + h(i) (y, t) ,
tbi
i = 1, . . . , l,
12
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
where tAi d(i) (y) collects the terms with smallest positive power of t and h(i) (y, t) collects
the remaining terms (whose powers of t exceed Ai ). Then f (i) (y) has Newton segment
Ii and its number of non-zero roots counted with multiplicities is |Ii |, the length of the
interval Ii .
Proposition 4.1. Assume that for any non-zero root ρ of f (i) , i = 1, . . . , l, either ρ is a
simple root of f (i) , or else d(i) (ρ) 6= 0. Then there exists t0 > 0 such for 0 < t < t0 , the
univariate polynomial ft (y) has only simple non-zero roots, with
X
r =
c(ρ)
non-zero real roots, where
of f (i) where

 1
0
c(ρ) =

2
the sum is over i = 1, . . . , l and then all non-zero real roots ρ
if the multiplicity m of ρ is odd,
if m is even and f (i) (y)/d(i) (ρ) > 0, for y near ρ,
if m is even and f (i) (y)/d(i) (ρ) < 0, for y near ρ.
In particular, if the non-zero roots of f (1) , . . . , f (l) are simple, then the number of nonzero real roots of ft for t > 0 small enough equals the total number of non-zero real roots
of f (1) , . . . , f (l) . This is the usual version of Viro’s theorem for univariate polynomials.
Proof. For each root ρ 6= 0 of f (i) (y) of multiplicity m, there will be m roots near ρ to
f (i) (y) + tAi d(i) (y) + h(i) (y, t) ,
for t > 0 sufficiently small. This gives |Ii | roots to ft (yt−ai )/tbi , and thus all solutions
to ft in C∗ , at least when t > 0 is sufficiently small. Indeed, let K ⊂ C∗ be a compact
set containing the non-zero roots of the facial polynomials f (1) (y), . . . , f (l) (y). Then, for
t > 0 sufficiently small, K contains the |Ii | roots to ft (yt−ai )/tbi that we just constructed.
The compact sets t−a1 K, . . . , t−al K are pairwise disjoint for t > 0 sufficiently small, and
this gives |I1 | + · · · + |Il | = d − p0 non-zero simple roots of ft for t > 0 small enough. But
this accounts for all the non-zero simple roots of ft .
We now determine how many roots of ft (yt−ai )/tbi are real. Roots close to ρ are real
only if ρ is real, and then the number of such real roots is determined by the first two
terms f (i) (y) + tAi d(i) (y) in t, as d(i) (ρ) 6= 0. But this polynomial has c(ρ) real roots near
ρ.
4.2. Upper bounds. We give upper bounds on the number of real roots of a generic
polynomial system with support a primitive near circuit
C = {0, ℓen , 2ℓen , . . . , kℓen , w1 , . . . , wn } .
As explained in Section 3, it suffices to bound the roots of a polynomial f of the form (3.5).
Consider the polynomial ft (y) depending on a real parameter t 6= 0 defined by
p
ν
Y
Y
N
ℓ λi
ft (y) := t · y
(gi (y )) −
(gi (y ℓ ))λi ,
i=1
i=p+1
where g1 , . . . , gν are generic polynomials of degree k. We will study how the number of
real roots of ft can vary as t runs from ∞ to 1. Note that f1 is our original eliminant f .
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
13
Remark 4.2. If p 6= 0 then ft (xn ) is the eliminant of the system
½ w
x 1 = t1/λ1 · g1 (xℓn ) ,
xwi = gi (xℓn ) i = 2, . . . , n .
If p = 0, then ft is t2 times the eliminant of the system
½ w
x 1 = t−1/λ1 · g1 (xℓn ) ,
xwi = gi (xℓn ) i = 2, . . . , . . . , n .
For any integer a, define a ∈ {0, 1, 2} by

 2 if a is positive and even
1 if a is positive and odd
a :=
 0 otherwise
A root ρ of a univariate polynomial is singular if it has multiplicity greater than 1.
Let χ(Y ) be the boolean truth value of Y , so that χ(0 > 1) = 0, but χ(0 < 1) = 1.
Proposition 4.3. The total sum of the multiplicities of the non-zero singular real roots
of ft for t ∈ R∗ is no more than 2kℓν − 2ℓ̄ (χ(δ = 0) + χ(N = 0)).
Moreover, if ℓ is even, so that N is odd due to the primitivity of C, then the total sum
of the multiplicities of the non-zero singular real roots of ft for t > 0 is equal to the
corresponding sum for t < 0. Hence, both numbers are no more 2kν.
Proof. Write ft = tF − G, where F and G are the two terms of f . Let ρ be a non-zero
root of ft for some t 6= 0. Then F (ρ)G(ρ) 6= 0, as the roots of g1 , . . . , gν are distinct. Note
that t = G(ρ)/F (ρ). Then ρ is a singular root of ft if and only if
(F ′ G − F G′ )(ρ) = 0 .
If N 6= 0, then the polynomial F ′ G − F G′ factors as
Ã
!
ν
Y
(4.2)
(F ′ G − F G′ )(y) = y N −1 (gi (y ℓ ))λi −1 · H(y)
i=1
where H is the polynomial defined by
H(y) =
ν
Y
i=1
with
D(z) =
p
X
i=1
N −1
¡
¢
gi (y ℓ ) · N + ℓy ℓ · D(y ℓ ) ,
ν
X
gi′ (z)
g ′ (z)
λi ·
−
.
λi · i
gi (z)
g
(z)
i
i=p+1
(If N = 0, then y
is replaced by y ℓ−1 , and the last factor in H is simply ℓD(y ℓ ).)
ℓ
Thus H(y) = h(y ), where h is a polynomial of degree kν − (χ(δ = 0) + χ(N = 0)) with
a non-zero constant term, as the gi are generic. If ρ is a non-zero singular real root of ft
for t 6= 0 then h(ρℓ ) = 0. Thus the total number of non-zero singular real roots of ft for
t ∈ R∗ is at most kℓν − ℓ̄ (χ(δ = 0) + χ(N = 0)).
14
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
A root ρ of ft has multiplicity m ≥ 2 for some t 6= 0 if and only if tF (i) (ρ) − G(i) (ρ) = 0
for i = 0, . . . , m − 1 and tF (m) (ρ) − G(m) (ρ) 6= 0. This is equivalent to the system
(4.3)
F (i) (ρ) · G(j) (ρ) − F (j) (ρ) · G(i) (ρ) = 0
F (i) (ρ) · G(j) (ρ) − F (j) (ρ) · G(i) (ρ) 6= 0
if 0 ≤ i, j ≤ m − 1,
if i = m and 0 ≤ j ≤ m − 1.
Solving (4.2) for H and using the expression of the kth derivative of F ′ G − F G′ as a sum
of polynomials of the form F (i) G(j) − F (j) G(i) , we can use (4.3) to deduce that ρ is a root
of multiplicity m ≥ 2 of ft for some t 6= 0 if and only if ρ is a root of multiplicity m−1
of H. Thus the total sum of the multiplicities of the non-zero singular real roots of ft for
t ∈ R∗ is
X
(4.4)
(mρ (H) + 1) ,
ρ a real root of H
where mρ (H) is the multiplicity of the root ρ of H. We see that (4.4) is bounded by
2kℓν − 2ℓ (χ(δ = 0) + χ(N = 0)) with equality when all roots of h are real and simple
(and positive if ℓ is even) so that the singular real roots of ft for t ∈ R∗ are real double
roots.
Finally, the statement concerning the case ℓ even and N odd is obvious after noting
that in this case the (non-zero) real roots of h(y ℓ ) come in pairs (ρ, −ρ), the function G/F
is an odd function, and N and δ are both odd integer numbers.
Remark 4.4. If {kℓen , w1 , . . . , wn } are affinely independent and N 6= 0, then the difference δ (3.6) of the degrees of the terms of f is non-zero, and the polynomials ft have the
same Newton segment for t 6= 0. Thus the number of real roots of ft can change only if t
passes through 0, or through a value c 6= 0 such that fc has a singular root.
If the difference δ = 0, then there is one number t∞ for which the degree of ft∞ drops.
If necessary, we may perturb coefficients of one gi so that the number of real roots of f
does not change, and the degree of ft∞ drops by one. This will result in no net change
in the number of real roots of ft as t passes through t∞ , for the root which ‘disappears’
in ft∞ is a real root at infinity. Similarly, if N = 0, then we may assume that there is
one number t0 for which the constant term of ft0 vanishes. Perturbing again if necessary
results in no net change in the number of non-zero real roots of ft as t passes through t0 .
Thus, the number of values c where the number of real roots of ft changes is finite by
Proposition 4.3, and hence it makes sense to define the numbers
r−∞ ,
r0− ,
r0+ ,
r+∞
as the numbers of real roots of ft as t tends to −∞, 0 by negative values, 0 by positive
values and +∞, respectively.
Recall that χ(Y ) denotes the boolean truth value of Y .
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
Proposition 4.5. We have
r0+ + r0−
≤ kℓ(ν − p) + χ(δ > 0)
2
r+∞ + r−∞
≤ kℓp + χ(N > 0) + χ(δ < 0)
2
!
à ν
X
|r0+ − r0− |
≤ kℓ
λi − kℓ(ν − p) + χ(δ > 0 is even)
2
i=p+1
!
à p
X
|r+∞ − r−∞ |
λi − kℓp + χ(N > 0 is even) + χ(δ < 0 is even)
≤ kℓ
2
i=1
15
Furthermore, if l is even and N is odd, we have
r0+ + r+∞
≤ kν + 1
2
Proof. As in the proof of Proposition 4.3, write ft = tF − G. We apply Proposition 4.1
(and its proof) to ft and f−t = −tF − G to estimate r0+ and r0− , respectively.
Let P be the common Newton polygon of ft (y) and f−t (y), as polynomials in y and t.
Projecting the lower faces of P onto the first (y) coordinate axis gives a single interval
I1 if δ ≤ 0, or the union of two intervals I1 and I2 if δ > 0. Here, I1 = [0, deg(G)], the
Newton segment of G and I2 = [deg(G), deg(F )], which has length δ. For both ft (y) and
f−t (y), the polynomial f (1) corresponding to I1 is just G. If δ > 0, then the polynomial
f (2) corresponding to I2 is the binomial ±MF − MG , which is the difference of the highest
degree terms of ±F and G.
Both binomials ±MF − MG have only simple non-zero roots. The polynomial f (1) = G
has singular roots if any of λp+1 , . . . , λν are not equal to 1. Since F and G have no common
root, the assumptions of Proposition 4.1 are fulfilled for both ft and f−t . The numbers
r0+ and r0− , hence r0+ + r0− and |r0+ − r0− |, are sums of contributions of the non-zero
real roots of G and of the non-zero real roots of ±MF − MG .
Consider contributions from roots ρ of G, which satisfy ρℓ = ζ, where ζ is a root of
some gi for p < i ≤ ν. This has multiplicity λi . If λi is odd, then ρ contributes 1 to
both r0± and hence 2 to r0+ + r0− and 0 to |r0+ − r0− |. If λi is even, then ρ contributes
2 or 0 to r0+ , depending upon the sign of G/F near ρ. Replacing t by −t, shows that it
contributes 2 or 0 to r0− , depending upon the sign of −G/F near ρ. Thus ρ contributes
2 both to r0+ + r0− and to |r0+ − r0− |.
Suppose now that δ > 0. Then each non-zero real root of the binomial ±MF − MG is
simple and thus contributes 1 to r0± . Both binomials have only one non-zero real root if
δ is odd. If δ is even, then MF − MG (resp. −MF − MG ) has 0 or 2 (resp. 2 or 0) real
roots according as the product of the coefficients of MF and MG is positive or negative.
It follows that the roots of these binomials contribute 2 to r0+ + r0− , and contribute 0 or
2 to |r0+ − r0− | according as δ is odd or even, respectively. Summing up all contributions
gives the desired upper bounds for r0+ + r0− and |r0+ − r0− |.
The upper bounds for r+∞ − r−∞ and |r+∞ − r−∞ | are obtained in exactly the same
way if we use the polynomial gt (y) = F (y) − tG(y) instead of ft (y). The Newton polygon
16
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Q of gt is the reflection of P in the horizontal line of height 1/2, so that the lower faces
of Q are the upper faces of P . They project to 1, 2, or 3 intervals on the first coordinate
axis, I1 = [0, N ] (if N 6= 0), I2 = [N, deg(F )], and I3 = [deg(F ), deg(G)], if δ < 0. The
polynomial f (2) corresponding to I2 is just F , and the other polynomials are binomials.
Finally, assume that ℓ is even, N is odd, and let us prove the last inequality. Using the
facts that the non-zero real roots of G come in pairs (ρ, −ρ), the function G/F is an odd
function, and that N and δ are both odd integers, we obtain r0+ ≤ 2k(ν − p) + χ(δ > 0).
Similarly, using the polynomial gt (y) = F (y) − tG(y) instead of ft (y), we obtain that
r+∞ ≤ 2kp + 1 + χ(δ < 0). Suming up these two inequalities gives the result.
Theorem 4.6. The number r of real solutions to a generic system with support the near
circuit C = {0, ℓen , 2ℓen , . . . , kℓen , w1 , . . . , wn } satisfies the following inequalities
³P
´
ν
r ≤ 2kℓp + kℓ
i=p+1 λi + χ(N > 0) + 1 − χ(δ > 0 is odd)
(4.5)
−χ(δ = 0) − ℓ (χ(δ = 0) + χ(N = 0)) ,
and
r ≤ 2kℓ(ν − p) + kℓ
(4.6)
¡Pp
i=1
¢
λi + χ(N > 0 is even) + 1 − χ(δ < 0 is odd)
−χ(δ = 0) − ℓ (χ(δ = 0) + χ(N = 0)) ,
where k, ℓ, N, ν, and λi are defined in Section 3.2.
Moreover, if ℓ is even and N is odd, then we have
(4.7)
r ≤ 2kν + 1
Remark 4.7. Since λ1 , . . . , λν are relatively prime, we see that the absolute upper bound
for such a near circuit with ℓ odd is
k(2ν − 1) + 2 ,
and this can be obtained if exactly one λi is odd, N > 0 and even, and δ is even and nonzero. This upper bound is maximized when {0, en , w1 , . . . , wn } forms a non degenerate
circuit, that is, if no proper subset is affinely dependent. If ℓ is even, the corresponding
absolute upper bound is 2kν + 1, and is also maximized when {0, en , w1 , . . . , wn } forms a
non degenerate circuit.
Proof of Theorem 4.6. Let f be the univariate eliminant of a generic polynomial system
with support C, which has the form (3.5). For an interval I ⊂ R, let ∆I be the (positive)
variation in the number of non-zero real roots of ft for t ∈ I. As noted in Remark 4.4, the
number of non-zero real roots of ft can change only if t passes through 0, or if t passes
through a value c 6= 0 such that fc has a real singular root. Passing through the value
t = 0, the variation of the number of real roots of ft is at most |r0+ − r0− |.
Recall that f = ft for t = 1 and that f has no singular roots. Considering the path
from t close to −∞ to t = 1, we obtain
r ≤ r−∞ + ∆(−∞,0) + |r0+ − r0− | + ∆(0,1) .
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
17
Considering the path from t close to +∞ to t = 1, we obtain
r ≤ r+∞ + ∆(1,+∞) .
Combining these two inequalities yields
r+∞ + r−∞ + |r0+ − r0− | ∆R∗
+
.
2
2
The number ∆R∗ is at most the total multiplicity of the singular real roots of ft for t 6= 0.
By Proposition 4.3, this is at most 2kℓν − 2ℓ (χ(δ = 0) + χ(N = 0)). The inequality (4.5)
follows then from (4.8) and Proposition 4.5.
Using the polynomial gt = F − tG in place of ft = tF − G, leads to
(4.8)
r ≤
r0+ + r0− + |r+∞ − r−∞ | ∆R∗
+
.
2
2
The inequality (4.6) is then obtained using Proposition 4.5.
Finally, considering the paths from t = 0 to t = 1, and from t close to +∞ to t = 1
gives
(4.9)
r ≤
r0+ + r+∞ ∆(0,+∞)
+
.
2
2
For ℓ even and N odd, the inequality (4.7) comes then from the corresponding statements
in Proposition 4.3 and Proposition 4.5.
(4.10)
r ≤
5. Constructions and sharp upper bounds
We now construct polynomials f having the form (3.5) with many real roots. In some
cases, this achieves the upper bound of Theorem 4.6 for the maximal number of real
solutions to generic systems with support a given near circuit.
Theorem 5.1. Let C be a primitive near circuit with k, ℓ, N , and λi as in Section 3.2.
Suppose that d1 , . . . , dν are nonnegative integers with di ≤ k such that
(5.1)
ℓ
ν
X
di λi < N + kℓ
p
X
λi .
i=1
i=1
If ℓ is odd, then there is a generic polynomial system with support C having
ν
X
di λi + d
i=1
real solutions, where d is the (positive) difference of the two sides of (5.1). If ℓ is even,
hence N is odd, then there is a generic polynomial system with support C having
2
ν
X
i=1
real solutions.
di + 1
18
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
We use Proposition 4.1 to determine the number of real roots for small t > 0 of a
polynomial of the form
Y
Y
(ζ − xℓ )m(ζ) ,
(t−b xℓ − ζ)m(ζ) − xµ
ft (x) = ta
ζ∈R2
ζ∈R1
where µ, a, b are positive integers, R1 and R2 are disjoint sets
P of positive real numbers,
and m(ζ) is a positive integer for ζ ∈ R1 ∪ R2 . Set µi := ℓ · ζ∈Ri m(ζ).
Lemma 5.2. Suppose that µ1 < aℓ/b < µ, and we have further that
(1) if ζ, ζ ′ ∈ R1 with m(ζ) even and m(ζ ′ ) odd, then ζ ′ < ζ, and
(2) if ζ, ζ ′ ∈ R2 with m(ζ) even and m(ζ ′ ) odd, then ζ ′ > ζ.
Let e (respectively o) be the number of ζ ∈ R1 ∪ R2 such that m(ζ) is even (respectively
odd). If ℓ is odd, then, for t > 0 sufficiently small, ft has exactly 2e + o + µ − µ1 simple
non-zero real roots. If ℓ is even and µ is odd, then, for t > 0 sufficiently small, ft has
exactly 2e + 2o + 1 simple non-zero real roots.
Proof. We use the notation of Proposition 4.1. The inequalities µ1 < aℓ/b < µ imply that
the lower hull of P consists of three segments whose projection onto the first coordinate
axis are the intervals I1 = [0, µ1 ], I2 = [µ1 , µ], and I3 = [µ, µ + µ2 ]. The corresponding
facial subpolynomials are
Y
(xℓ − ζ)m(ζ)
f (1) (x) =
ζ∈R1
f (2) (x) = xµ1 −
f (3) (x) = −xµ
¡Y
Y
ζ∈R2
¢
ζ m(ζ) · xµ
(ζ − xℓ )m(ζ) .
ζ∈R2
Q
Note that ζ∈R2 ζ m(ζ) > 0 since R2 consists of positive real numbers. Thus the non-zero
roots of the binomial f (2) are simple, with µ − µ1 of them real. Proposition 4.1 applies as
we can see from the expansions (4.1) for f (1) and f (3) .
Y
bµ
b
ft (xt ℓ )/ta = f (1) (x) − t ℓ −a xµ
(ζ − xℓ tb )m(ζ)
ζ∈R2
= f
(1)
(x) − t
bµ
−a
ℓ
ft (x) = f (3) (x) + ta
µ
x (
Y
Y
ζ m(ζ) ) + h(1) (x, t) .
ζ∈R2
(t−b xℓ − ζ)m(ζ)
ζ∈R1
= f
(3)
a−
(x) + t
bµ1
ℓ
xµ1 + h(3) (x, t) .
Assume that ℓ is odd. Then the map x 7→ xℓ is a bijection from the real roots of f (1)
(resp., f (3) ) to R1 (resp., R2 ). Conditions (1) and (2) imply that the contribution of any
real root ρ such that ρℓ = ζ ∈ Ri and m(ζ) is even is equal to 2.
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
19
Assume that ℓ is even and µ is odd. Then the real roots of f (1) (resp., f (3) ) come in
pairs (ρ, −ρ) with ρℓ = ζ ∈ R1 (resp., R2 ). If m(ζ) is odd, then the contributions of ρ and
−ρ are both equal to 1. If m(ζ) is even, then one contribution is 2, while the other is 0,
as µ is odd. Finally, note that µ − µ1 = 1 if ℓ is even and µ is odd.
Proof of Theorem 5.1. Set
p
X
(k − di )λi .
µ := N + ℓ
i=1
The inequality (5.1) can be rewritten as
ℓ
ν
X
di λi < µ .
i=p+1
Then, by Lemma 5.2 there exist polynomials h1 , . . . , hn with distinct roots such that hi
has degree di with di real roots and the polynomial
p
ν
Y
Y
ℓ λi
(hi (x )) −
(hi (xℓ ))λi
g(x) := x
µ
i=1
i=p+1
has either
ν
X
di λi + d
i=1
or else
2
ν
X
di + 1
i=1
simple real roots according as ℓ is odd, or ℓ is even and N is odd, respectively. The
polynomial g(x) can be rewritten as
g(x) = xN
p
Y
¡
i=1
ν
Y
¢λ
¡
¢λ
xℓ(k−di ) hi (xℓ ) i −
hi (x)ℓ i
i=p+1
If di = k, set gi (x) := hi (x). Otherwise, set
gi (x) := ǫ(1 + x + · · · + xk−di −1 ) + xk−di hi (x)
gi (x) := hi (x) + ǫ(x
di +1
k
+ ··· + x )
1≤i≤p
p+1≤i≤ν.
For sufficiently small ǫ > 0, the polynomial
f (x) = x
N
p
Y
¡
i=1
ν
Y
¢λ
¡
¢λ
gi (xℓ ) i −
gi (xℓ ) i
i=p+1
has simple roots and at least the same number of real roots as g.
20
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
P
Theorem 5.3. Assume that N > kℓ νi=p+1 λi and let m be the maximal number of real
solutions to a generic system with support the near circuit C.
If ℓ is even, then m = 2kν + 1.
Suppose now that ℓ is odd.
(1) If λ1 , . . . , λp are even, then
m = 2kp + k
ν
X
λi + δ .
i=p+1
(2) If exactly one number among λ1 , . . . , λp is odd, k = ℓ = 1 and δ is odd, then
ν
X
m = 2p + 1 +
λi .
i=p+1
(3) If λp+1 , . . . , λn are even, then
m = 2k(n − p) + k
p
X
λi + N .
i=1
(4) If exactly one number among λp+1 , . . . , λn is odd, k = ℓ = 1 and N is odd, then
m = 2(n − p) + 1 +
p
X
λi .
i=1
Proof. We apply Theorem 5.1 with each di = k. The case of ℓ even is a direct consequence
of Theorem 4.6 and Theorem 5.1. Suppose now that ℓ is odd and let B1 and B2 be the
upper bounds for the number of real solutions to a generic system with support C which
are given in Theorem 4.6 by formulas (4.5) and (4.6), respectively. Set
B− := k
ν
X
λi + d ,
i=1
P
where d := N − kℓ νi=p+1 λi > 0. Note that δ, N ≥ d, so δ, N > 0. By Theorem 5.1, the
number B− is a lower bound on the maximal number of real solutions of a generic system
with support C.
We check that B1 ≥ B− and analyze the conditions under which B1 = B− . As δ > 0
and N > 0, we have
p
X
(2 − λi ) + δ − d .
B1 − B− = k
i=1
We have



P
0 if kℓ pi=1 λi is even
P
1 if kℓ pi=1 λi is odd and d is odd
δ−d =

 −1 if kℓ Pp λ is odd and d is even
i=1 i
If δ − d = 0, then B1 ≥ B− with equality only if λ1 , . . . , λp are even. This proves Part (1).
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
21
Pp
If δ − d = 1, then B1 > B− . Assume now that δ − d = −1. Then kℓ i=1 λi is odd,
P
P
d is even and B1 − B− = kℓ pi=1 (2 − λi ) − 1. Since kℓ pi=1 λi is odd, kℓ is odd and at
least one number among λ1 , . . . , λp is odd. Thus B1 − B− ≥ kℓ − 1 with equality only if
exactly one number among λ1 , . . . , λp is odd. Part (2) now follows.
Parts (3) and (4) are similar.
Theorem 5.4. Assume that λi ∈ {1, 2} for i = 1, . . . , ν, ℓ is odd, and let m be the
maximal number of real solutions to a generic system with support the near circuit C.
P
(1) If N > kℓ νi=p+1 λi , then
m = k
ν
X
λi + N − kℓ
i=1
(2) Suppose that ℓ = 1. If N < k
Pν
i=p+1
(
m = v(C) = max k
Pν
i=p+1
λi .
λi , then
ν
X
λi , N + k
p
X
i=1
i=p+1
λi
)
.
Proof. For Part (1), the number m equals the upper bound given by Descartes’s rule of
signs when applied to a polynomial of the form (3.5). Theorem 5.1 with each di = k
implies the existence of a polynomial with
form with m real roots.
Pthis
P
ν
For Part (2), we have that N < k i=p+1 λi . If we also have N + k pi=1 λi >
Pν
k i=p+1 λi , then there exist nonnegative integers d1 , . . . , dν ≤ k such that
!
à ν
p
X
X
λi di ∈ {1, 2} ,
λi −
d = N +k
i=1
i=1
as λi ∈ {1, 2}. By Theorem
5.1, there exists a polynomial of the form (3.5) having
Pp
P
ν
λ
non-zero real
λ
d
+
d
=
N
+
k
i
i
i
i=1
i=1
P
Proots.
p
Finally, suppose that N + k i=1 λi ≤ k νi=p+1 λi . Consider a polynomial
p
ν
Y
Y
λi
(gi (x)) −
(gi (x))λi ,
ft (x) = t · x
N
i=1
i=p+1
where g1 , . . . , gν are polynomials of degree k with non-zero constant terms. TheP
lower part
of the Newton polygon of ft (x) consists of a single segment projecting onto [0, k νi=p+1 λi ].
Hence, if Proposition 4.1 applies, the number of real roots of ft for t > 0 small enough is
the sum of contributions of the non-zero real roots of gp+1 , . . . , gν . Choosing polynomials
g1 , . . . , gν with distinct roots satisfying conditions
(1) If i, j > p, then gi has k positive roots, and if λi is odd and λj is even, then the
leading coefficient of gi is positive and every root of gi is less than every root of gj .
(2) The polynomials g1 , . . . , gp are positive at each root of of gp+1 , . . . , gν .
P
By Proposition 4.1, ft has k νi=p+1 λi non-zero real roots for t > 0 small enough.
22
BENOÎT BERTRAND, FRÉDÉRIC BIHAN, AND FRANK SOTTILE
Remark 5.5. The example of Section 2 is a special case of Part(1) of Theorem 5.4.
Indeed, in Section 2, we have
X
w0 = en , wi = ei i = 1, . . . , n−1, and wn =
ǫi ei + len ,
so that
N = l, ν = |ǫ| + 1 =
X
λi , and p = |ǫ| .
i
Then the maximum number of Part(1) of Theorem 5.4 is
m = k(|ǫ| + 1) + l − k ,
which is what we found in Section 1.
Theorem 5.6. The number of real roots of a generic system with support a primitive
near circuit C = {0, ℓen , . . . , kℓen , w1 , . . . , wn } in Rn is at most (2ν − 1)k + 2 if ℓ is odd,
or 2kν + 1 if ℓ is even. Moreover, these bounds are sharp.
Proof. As the numbers λ1 , . . . , λν are coprime, at least one is odd. Since λj ≤ 2, the
upper bound for odd ℓ follows from Theorem 4.6 (see Remark 4.7). The sharpness of this
bound follows from Theorem 5.4 for a primitive near circuit with
P all λ1 , . . . , λν but one
equal to 2 and one which is equal to 1, and where and N − kℓ νi=p+1 λi is positive and
even (Example 3.6 shows that such a near circuit exists).
The bound 2kν + 1 for ℓ even comes from Theorem 4.6, its sharpness follows from
Theorem 5.3.
A near circuit with k = 1 is just a circuit. The following result is a particular case of
the previous one.
Theorem 5.7. The number of real roots of a generic system with support a primitive
circuit C = {0, ℓen , , w1 , . . . , wn } in Rn is at most 2ν + 1, and this bound this sharp.
The absolute upper bound for the number of real roots of a generic system with support
a primitive circuit in Rn is 2n + 1, and this bound is sharp. Moreover, this bound can be
attained only for non-degenerate circuits.
Proof. We only need to prove the last sentence as the others are corollaries of Theorem 5.6.
For this, we note that the bound 2ν + 1 is obtained when δ and N are non-zero. Hence
the absolute bound 2n + 1 is obtained when ν = n, δ and N are non-zero, which is exactly
the case of a non-degenerate circuit.
References
[1] Saugata Basu, Richard Pollack, and Marie-Françoise Roy, Algorithms in real algebraic geometry,
Algorithms and Computation in Mathematics, vol. 10, Springer-Verlag, Berlin, 2003.
[2] Frederic Bihan, Viro method for the construction of real complete intersections, Advances in Mathematics, vol. 169, No. 2, (2002), 177–186.
[3] René Descartes, Géométrie. (1636) In: A source book in Mathematics. Massachussetts, Harvard
University Press 1969, 90–93.
[4] Robin Hartshorne, Algebraic geometry, Springer-Verlag, New York, 1977, Graduate Texts in Mathematics, No. 52.
[5] Askold G. Khovanskii, Fewnomials, Trans. of Math. Monographs, 88, AMS, 1991.
POLYNOMIAL SYSTEMS WITH FEW REAL ZEROES
23
[6] Anatoli G. Kouchnirenko, A Newton polyhedron and the number of solutions of a system of k equations
in k unknowns, Usp. Math. Nauk. 30 (1975), 266–267.
[7] Tien-Yien Li, J. Maurice Rojas, and Xiaoshen Wang, Counting real connected components of trinomial
curve intersections and m-nomial hypersurfaces, Discrete Comput. Geom. 30 (2003), no. 3, 379–414.
[8] Bernd Sturmfels, On the number of real roots of a sparse polynomial system, Hamiltonian and gradient flows, algorithms and control, Fields Inst. Commun., vol. 3, American Mathematical Society,
Providence, 1994, pp. 137–143.
[9] Bernd Sturmfels, Gröbner bases and convex polytopes, American Mathematical Society, Providence,
RI, 1996.
[10] Oleg Viro. Gluing of algebraic hypersurfaces, smoothing of singularities and construction of curves.(in
russian). Proc. Leningrad Int. Topological Conf.,Leningrad, 1982, Nauka, Leningrad, pages 149–197,
1983.
. Gluing of plane algebraic curves and construction of curves of degree 6 and 7. Lecture Notes
[11]
in Mathematics 1060, pages 187–200, 1984.
Université de Genève, Section de mathematiques, 2-4, rue du Lièvre, Case postale 64,
1211 Genève 4, Suisse
E-mail address: benoit.bertrand@math.unige.ch
Laboratoire de Mathématiques, Université de Savoie, 73376 Le Bourget-du-Lac Cedex,
France
E-mail address: Frederic.Bihan@univ-savoie.fr
Department of Mathematics, Texas A&M University, College Station, TX 77843, USA
E-mail address: sottile@math.tamu.edu
URL: http://www.math.tamu.edu/~sottile
Download